What is AI actually worth to your team?
Honest numbers in about two minutes. Built on peer-reviewed research and our own client work, with every assumption on show. No email needed to see your results.
The calculator
Three steps. Your numbers, not ours.
Describe your team, the time AI could save, and how you would roll it out. The maths runs in your browser, and every assumption sits in plain sight under your results.
- 1Your team
- 2Time saved
- 3Your rollout
Step 1 of 3
Your team
Headcount and a rough salary mix. Round numbers are fine.
How many people in each salary band?
Step 2 of 3
Time saved
How many hours could AI hand back to each person, each day? The markers show what the research found.
Step 3 of 3
Your rollout
How you roll AI out sets how many people actually use it, and how fast they get there. Pick the closest match.
Fine-tune the assumptions
Salary band midpoints
Our weighting, informed by research showing AI lifts less-experienced staff most. In a 2023 Harvard and BCG experiment with 758 consultants, below-average performers improved 43% with AI against 17% for above-average. And a 2025 study of 5,172 customer support agents published in the Quarterly Journal of Economics found around 30% gains for the least experienced and minimal impact for the most experienced.
Your results
What AI is worth to your team
10.4% of your team’s working time
The honest bit
Capacity is not the same as cash
Only 39% of surveyed executives in McKinsey’s 2025 State of AI survey attribute any EBIT impact to AI. Most saved time never reaches the P&L, so we show both numbers.
The ramp
Value builds month by month. Waiting has a cost.
Cumulative captured value over 18 months, structured programme ramp.
Waiting six months costs about £136,000 in captured value and 9,984 hours.
In BCG’s 2025 survey of 1,250 executives, the 5% of firms furthest ahead on AI reported around five times the revenue gains of the 60% majority. The gap is widening.
That is about 4.2 hours a week back, per person.
The quality dividend
The work gets better, not just faster
In a randomised study published in Science in 2023, professionals using AI completed writing tasks in 40% less time, with 18% higher quality.
In the Harvard and BCG experiment, output quality on tasks within AI’s capabilities rose by around 30 to 34%, the peer-reviewed figures.
Our assumptions and sources
| Working month | 20 working days a month. |
|---|---|
| Working day | 8 hours. |
| Salary band midpoints | 37,500 / 75,000 / 125,000. Deft defaults, editable in the fine-tune panel. |
| Band uplift weights | 1.2 / 1.0 / 0.8 across the three salary bands. A stated Deft assumption, informed by the research below, not a study output. |
| Capture rate | 40% by default. Adjustable from 10 to 100%. |
| Adoption | Headline numbers use the flat rollout-ceiling adoption. The chart uses the month-by-month ramp. |
Sources
- Dell’Acqua et al., the Harvard and BCG experiment with 758 consultants. Organization Science, 2026
- Noy and Zhang, randomised writing-task study with 453 professionals. Science, 2023
- Brynjolfsson, Li and Raymond, study of 5,172 customer support agents. Quarterly Journal of Economics, 2025
- McKinsey, survey of executives worldwide. The State of AI, 2025
- BCG, survey of 1,250 executives. The Widening AI Value Gap, 2025
- Microsoft WorkLab, self-reported findings from Copilot’s earliest users. Work Trend Index, 2023
Keep these numbers
Take your numbers with you
Enter your email to unlock a print-ready summary of your results, built from your inputs. Yours to keep or share.
One unlock, right here on the page. Nothing lands in your inbox unless you tick the box.
Done. Your summary is ready. Save it as a PDF or send it to print to keep a copy of these numbers.
Where Deft comes in
Saving the time is step one. Capturing it is the work.
High performers are nearly three times more likely to have fundamentally redesigned workflows, per McKinsey’s 2025 State of AI survey.
The gap between using AI and getting value from it is enablement. That is what we do.
Start a conversation